Bug in pthread_setcancelstate()?

[Erik Hensema]

Hi,

>From the pthread_setcancelstate() from SunOS 5.5.1:

 pthread_setcancelstate
     pthread_setcancelstate()  automically   sets   the   calling
     thread's  cancellation  state to the specified state and, if
     oldstate is not NULL, stores the previous cancellation state
     in oldstate.

In cancel.c:

 if (((self = pthread_self ()) != NULL) &&
     (state == PTHREAD_CANCEL_ENABLE ||
      state == PTHREAD_CANCEL_DISABLE))
   {

     *oldstate = self->cancelState;     <---- crash if oldstate==NULL
     self->cancelState = state;
     result = 0;

   }

As you can see, it will crash if oldstate==NULL.

The obvious fix is to add the line " if(oldstate!=NULL)" above the
"*oldstate=".

ps. I haven't checked this with the Posix standard, so it can also be a
deviation from the standard by Sun; however, I think the code is more robust
this way.

-- 
Erik Hensema
Work: erik.hensema@group2000.nl
Home: erik@hensema.xs4all.nl